Filters








1,928 Hits in 4.4 sec

Software-defined Design Space Exploration for an Efficient DNN Accelerator Architecture [article]

Ye Yu, Yingmin Li, Shuai Che, Niraj K. Jha, Weifeng Zhang
2020 arXiv   pre-print
We demonstrate that it can be efficaciously used in application-driven accelerator architecture design.  ...  In this article, we propose an application-driven framework for architectural design space exploration of DNN accelerators.  ...  It searches for the best layers on CIFAR-10 [47] and transfers the architectures to ImageNet [1] for object detection.  ... 
arXiv:1903.07676v2 fatcat:kgy2xwylwzdbdjlpjraku5n6ce

Automated design of error-resilient and hardware-efficient deep neural networks [article]

Christoph Schorn, Thomas Elsken, Sebastian Vogel, Armin Runge, Andre Guntoro, Gerd Ascheid
2019 arXiv   pre-print
It is thus desirable to exploit optimization potential for error resilience and efficiency also at the algorithmic side, e.g., by optimizing the architecture of the DNN.  ...  Since there are numerous design choices for the architecture of DNNs, with partially opposing effects on the preferred characteristics (such as small error rates at low latency), multi-objective optimization  ...  Energy consumption of DNN accelerators is dominated by data transfers to and from memory [88] .  ... 
arXiv:1909.13844v1 fatcat:x4pdboypqjepbnss4papdgt7ce

AlphaX: eXploring Neural Architectures with Deep Neural Networks and Monte Carlo Tree Search [article]

Linnan Wang, Yiyang Zhao, Yuu Jinnai, Yuandong Tian, Rodrigo Fonseca
2019 arXiv   pre-print
AlphaX improves the search efficiency by adaptively balancing the exploration and exploitation at the state level, and by a Meta-Deep Neural Network (DNN) to predict network accuracies for biasing the  ...  Finally, we show the searched architecture improves a variety of vision applications from Neural Style Transfer, to Image Captioning and Object Detection.  ...  Furthermore, while prior works on MCTS applied to Architecture Search [40, 28] only report performance on CIFAR-10, our distributed Al-phaX system can scale up to 32 machines (with 32 acceleration ratio  ... 
arXiv:1903.11059v2 fatcat:w5ecvttlyzebthmsqmepw4dy3q

NASCaps: A Framework for Neural Architecture Search to Optimize the Accuracy and Hardware Efficiency of Convolutional Capsule Networks [article]

Alberto Marchisio, Andrea Massa, Vojtech Mrazek, Beatrice Bussolino, Maurizio Martina, Muhammad Shafique
2020 arXiv   pre-print
Moreover, due to their extreme computational and memory requirements, DNNs are employed using the specialized hardware accelerators in IoT-Edge/CPS devices.  ...  Recent studies have shown that powerful methods to automatically select the best/optimal DNN model configuration for a given set of applications and a training dataset are based on the Neural Architecture  ...  Transferability of the Selected DNNs Across Different Datasets.  ... 
arXiv:2008.08476v1 fatcat:ziy6ni6wtvb2fgwdxzdcmxypnq

ENOS: Energy-Aware Network Operator Search for Hybrid Digital and Compute-in-Memory DNN Accelerators [article]

Shamma Nasrin, Ahish Shylendra, Yuti Kadakia, Nick Iliev, Wilfred Gomes, Theja Tulabandhula, Amit Ranjan Trivedi
2021 arXiv   pre-print
This work proposes a novel Energy-Aware Network Operator Search (ENOS) approach to address the energy-accuracy trade-offs of a deep neural network (DNN) accelerator.  ...  The search in ENOS is formulated as a continuous optimization problem, solvable using typical gradient descent methods, thereby scalable to larger DNNs with minimal increase in training cost.  ...  Among recent NAS approaches, NASnet [1] learns the best convolutional architecture on a smaller dataset using reinforcement learning and then transfers the learned architecture to a larger dataset.  ... 
arXiv:2104.05217v1 fatcat:aw66hzjb6je3peanwot7babewy

Domain-specific Genetic Algorithm for Multi-tenant DNNAccelerator Scheduling [article]

Sheng-Chun Kao, Tushar Krishna
2021 arXiv   pre-print
This work looks at the problem of supporting multi-tenancy on such accelerators. In particular, we focus on the problem of mapping layers from several DNNs simultaneously on an accelerator.  ...  Given the extremely large search space, we formulate the search as an optimization problem and develop a specialized genetic algorithm called G# withcustom operators to enable structured sample-efficient  ...  Sub-Accelerator Architecture and Schedule Each sub-accelerator in our system is a conventional DNN accelerator that is comprised of an array of Processing Elements (PE).  ... 
arXiv:2104.13997v2 fatcat:czzuamwzy5ee3kxno2rs7czi5e

AccSS3D: Accelerator for Spatially Sparse 3D DNNs [article]

Om Ji Omer, Prashant Laddha, Gurpreet S Kalsi, Anirud Thyagharajan, Kamlesh R Pillai, Abhimanyu Kulkarni, Anbang Yao, Yurong Chen, Sreenivas Subramoney
2020 arXiv   pre-print
The SSpNNA accelerator core together with the 64 KB of L1 memory requires 0.92 mm2 of area in 16nm process at 1 GHz.  ...  We present ACCELERATOR FOR SPATIALLY SPARSE 3D DNNs (AccSS3D), the first end-to-end solution for accelerating 3D scene understanding by exploiting the ample spatial sparsity.  ...  SCALE-UP ARCHITECTURE FOR 3D SPATIAL SPARSITY In this section, we present ACCELERATOR FOR SPATIALLY SPARSE 3D DNNs, an architecture ( Figure 13 ) targeting 3D spatially sparse DNN applications. A.  ... 
arXiv:2011.12669v1 fatcat:vhtzlpqvh5ciri2mzmza3cjxqi

Neural Architecture Search using Deep Neural Networks and Monte Carlo Tree Search [article]

Linnan Wang, Yiyang Zhao, Yuu Jinnai, Yuandong Tian, Rodrigo Fonseca
2019 arXiv   pre-print
AlphaX improves the search efficiency by adaptively balancing the exploration and exploitation at the state level, and by a Meta-Deep Neural Network (DNN) to predict network accuracies for biasing the  ...  Finally, we show the searched architecture improves a variety of vision applications from Neural Style Transfer, to Image Captioning and Object Detection.  ...  Inspired by this idea, we present AlphaX, a NAS agent that uses MCTS for efficient architecture search with Meta-DNN as a predictive model to estimate the accuracy of a sampled architecture.  ... 
arXiv:1805.07440v5 fatcat:nd6o3uhplvgfrflyiddztz6a2e

A Scalable System-on-Chip Acceleration for Deep Neural Networks

Faisal Shehzad, Muhammad Rashid, Mohammed H Sinky, Saud S Alotaibi, Muhammad Yousuf Irfan Zia
2021 IEEE Access  
The work in [9] presents a deep learning accelerator unit (DLAU) which is a scalable accelerator architecture for large-scale DNN using FPGA.  ...  for the transfer of weights and input data samples.  ... 
doi:10.1109/access.2021.3094675 fatcat:5qiye7xeszhv5b3al2uhq6ib5y

Efficient Visual Recognition with Deep Neural Networks: A Survey on Recent Advances and New Directions [article]

Yang Wu, Dingheng Wang, Xiaotong Lu, Fan Yang, Guoqi Li, Weisheng Dong, Jianbo Shi
2021 arXiv   pre-print
Deep neural networks (DNNs) have largely boosted their performances on many concrete tasks, with the help of large amounts of training data and new powerful computation resources.  ...  In this paper, we present the review of the recent advances with our suggestions on the new possible directions towards improving the efficiency of DNN-related visual recognition approaches.  ...  The second way is to yield the searched architecture to inherit the weights which are produced in previous searched architectures [186] , [191] , [198] .  ... 
arXiv:2108.13055v2 fatcat:nf3lymdbvzgl7otl7gjkk5qitq

Accelerating DNN-based 3D point cloud processing for mobile computing

Bosheng Liu, Xiaoming Chen, Yinhe Han, Jiajun Li, Haobo Xu, Xiaowei Li
2019 Science China Information Sciences  
We present the first accelerator architecture that dynamically configures the hardware onthe-fly to match the computation of both neighbor point search and MAC computation for point-based DNNs.  ...  Keywords deep neural network acceleration, point cloud data, neighbor point search, mobile robotics, hardware architecture Citation Liu B S, Chen X M, Han Y H, et al.  ...  This further confirms that accelerating neighbor point search is critical for point-based DNNs as the scale of point cloud data increases from application to application.  ... 
doi:10.1007/s11432-019-9932-3 fatcat:py7ttpbxunfuxfyprv2g2bpkse

Neural Architecture Search Using Deep Neural Networks and Monte Carlo Tree Search

Linnan Wang, Yiyang Zhao, Yuu Jinnai, Yuandong Tian, Rodrigo Fonseca
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
AlphaX improves the search efficiency by adaptively balancing the exploration and exploitation at the state level, and by a Meta-Deep Neural Network (DNN) to predict network accuracies for biasing the  ...  Finally, we show the searched architecture improves a variety of vision applications from Neural Style Transfer, to Image Captioning and Object Detection.  ...  Appendix The appendix is available at https://arxiv.org/abs/1805.07440.  ... 
doi:10.1609/aaai.v34i06.6554 fatcat:yiaje77muvcg3pa4dcxv4rjiju

Bringing AI To Edge: From Deep Learning's Perspective [article]

Di Liu, Hao Kong, Xiangzhong Luo, Weichen Liu, Ravi Subramaniam
2020 arXiv   pre-print
search.  ...  To bridge the gap, a plethora of deep learning techniques and optimization methods are proposed in the past years: light-weight deep learning models, network compression, and efficient neural architecture  ...  At the same time, as we are witnessing the shift of DNN designs from manual design to automatic search, i.e., neural architecture search, which has demonstrated its capability to design more accurate DNNs  ... 
arXiv:2011.14808v1 fatcat:g6ib7v7cxbdglihkizw5ldsxcu

Guided Sampling-based Evolutionary Deep Neural Network for Intelligent Fault Diagnosis [article]

Arun K. Sharma, Nishchal K. Verma
2021 arXiv   pre-print
In this paper, we have proposed a novel framework of evolutionary deep neural network which uses policy gradient to guide the evolution of DNN architecture towards maximum diagnostic accuracy.  ...  We have formulated a policy gradient-based controller which generates an action to sample the new model architecture at every generation.  ...  If the best model at a generation is transferred for initialization of the DNN weight matrices in the next generation, it makes the training and evaluation of the models faster.  ... 
arXiv:2111.06885v1 fatcat:wcf6ojcpubeidduxttx6b3w7la

DNNExplorer: A Framework for Modeling and Exploring a Novel Paradigm of FPGA-based DNN Accelerator [article]

Xiaofan Zhang, Hanchen Ye, Junsong Wang, Yonghua Lin, Jinjun Xiong, Wen-mei Hwu, Deming Chen
2021 arXiv   pre-print
under the proposed paradigm and deliver optimized accelerator architectures for existing and emerging DNN networks.  ...  optimized accelerator architectures following the proposed paradigm by simultaneously considering both FPGAs' computation and memory resources and DNN networks' layer-wise characteristics and overall  ...  To capture the runtime of searching the optimal architecture given input model and hardware constraints, we perform five independent searches using Intel i5-650 CPU for each case and list the average search  ... 
arXiv:2008.12745v2 fatcat:wyuohljc6rgmzoe2muahbpt6iq
« Previous Showing results 1 — 15 out of 1,928 results